AI agents now book, pay, approve and execute tasks on people’s behalf. Most AI agent identity verification confirms which account an agent is using. It rarely confirms which human authorised it.
This white paper sets out multi-factor delegation: a way to bind a verified human to the authority an agent holds. It explains the gap, the pattern that closes it and how the identity industry could put it into practice.
The Problem
Why AI Agent Identity Verification Stops at the Account
Today an agent usually connects to its human through an API key or an account session. That is possession alone. If the key is stolen, the agent keeps running. Every system downstream keeps treating it as authorised.
Passwords once had the same weakness. The fix was a second factor of a different type, something only the real person could produce. Multi-factor delegation applies the same idea to AI agents. It pairs the key an agent holds with a live check of the human who authorised it.
Passkeys and authenticated sessions help, but they don’t close the gap on their own. They prove a device or a login. They don’t prove who is authorising a specific agent action.
The factor is not on the agent. It is on the delegation.
Three Questions Every Agent Action Raises
Every action an agent takes raises three questions:
- Who is the agent? Agent identity platforms and protocols already answer this.
- What may it do? Permissions and governance policy set this today.
- Who authorised it? Usually inferred from account access. Multi-factor delegation is designed to prove it directly.
That doesn’t mean a face scan every time an agent does something. Checks run at policy-driven moments, when the risk warrants it. The paper explains how a signed “human present” attestation works and when each level of check applies.
What You’ll Learn
What’s Inside the White Paper
- The accountability gap. Why single-factor delegation is becoming the shared-password problem for AI agents.
- A precise definition of “human present”. What it covers and how it differs from continuous biometric monitoring.
- One delegation, step by step. A five-step walkthrough from verification to audit trail.
- Presence and attribution. Three levels of assurance, from anonymous but unique to full verified identity, and when each fits.
- The agent trust fabric, layer by layer. Where trust is bound today across identity platforms, protocols, payment rails, open source and accountability.
- Five plays in sequence. How the human factor could move from integration to standards, payments, open source and evidence.
- The competitive landscape. Who is working on the human factor today and where they overlap.
- Practical FAQs. Privacy, fallback routes, latency and deployment options.
Why This Matters
Why Delegation is Becoming an Accountability Question
Agent identity is being built quickly. Identity platforms, protocol groups and startups are all issuing agent credentials and permissions, and acquisitions in the space accelerated through 2025 and 2026.
Liability is likely to land on whoever deploys the agent. That party will need evidence of which human authorised what. Insurers writing AI liability cover and regulation such as the EU AI Act may play the role that PCI compliance and cyber insurance played for MFA. The paper sets out the evidence and what it means for teams deploying agents.
Who It’s For
Who Should Read It
The paper is written for teams responsible for AI agent identity verification, governance and risk:
- Identity and access leaders deploying AI agents across the enterprise
- Product teams building agent platforms, protocols or digital wallets
- Payments and fraud teams preparing for agentic commerce
- Risk, compliance and security teams accountable for what agents do
- Insurers, auditors and standards contributors working on AI accountability
Download Paper
FAQs
What is AI agent identity verification?
AI agent identity verification establishes which agent is acting and which account or organisation it represents. Multi-factor delegation adds a second question: which verified human authorised the delegation.
What is multi-factor delegation?
It pairs the credential an agent holds with a live check of the human who authorised it. The check happens when authority is granted and again at higher-risk moments.
Does it mean constant face scans?
No. Checks run at the moments policy requires, not on every action. Often an existing credential is enough.
Where does VerifEye fit?
VerifEye proposes to supply the human factor: a privacy-preserving face check that sits inside the platforms, protocols and payment rails agents already use. The face-verification engine behind it runs more than a billion verifications a day. Its application to agent delegation is at proof-of-concept stage.
Contents:
Share this: